Arkansas Tech University

MATH 2934: Calculus III

Dr. Marcel B. Finan

13.8 Extrema of Functions in Two Variables

Just like functions of a single variable, functions of several variables can have
local and global extrema, i.e., local and global maxima and minima.
We say that f (x, y) has a global maximum at a point (a, b) of its domain
Df if f (x, y) ≤ f (a, b) for all points (x, y) in Df . That is, f (a, b) is the largest
value of f in Df .
We say that f (x, y) has a global minimum at a point (a, b) of its domain
Df if f (a, b) ≤ f (x, y) for all points (x, y) in Df . That is, f (a, b) is the small-
est value of f in Df .
We say that f (x, y) has a local maximum at a point (a, b) of its domain
Df if there is an R > 0 such that f (x, y) ≤ f (a, b) for all points (x, y) in
Df satisfying (x − a)2 + (y − b)2 < R2 .
We say that f (x, y) has a local minimum at a point (a, b) of its domain
Df if there is an R > 0 such that f (a, b) ≤ f (x, y) for all points (x, y) in
Df satisfying (x − a)2 + (y − b)2 < R2 .
Collectively, local maxima and local minima are called local extrema. Sim-
ilar definition for global extrema. Figure 13.8.1 provides an example of a
function with local and global extrema.

Figure 13.8.1

Recall that for single-variable functions y = f (x), if x = c is a local maximum

or a minimum point, then either f 0 (c) = 0 or f 0 (c) does not exist. A point
(c, f (c)) such f 0 (c) = 0 or f 0 (c) does not exist is called a critical point.
Thus, the recipe for finding a maximum or a minimum point is to locate
critical points. Something similar happens for functions of two variables.
Points where the gradient is either zero or undefined are called critical
points of a function f (x, y).
We next show that local extrema are critical points.

1
Theorem 13.8.1
If f has a local maximum or a local minimum at a non-boundary point (a, b)
in its domain then ∇f (a, b) = ~0. That is, (a, b) is a critical point.

Proof.
Suppose f has a local extremum at a point (a, b). Define g(x) = f (x, b).
Then g(x) has a local extremum at x = a so that g 0 (a) = 0 = fx (a, b).
Likewise, the function G(y) = f (a, y) has a local extremum at y = b so that
G0 (b) = 0 = fy (a, b). Hence, ∇f (a, b) = ~0

Recall that the tangent plane to the surface z = f (x, y) is z = f (a, b) +

fx (a, b)(x − a) + fy (a, b)(y − b) so from the above theorem this equation
becomes z = f (a, b). That is, the tangent plane at a local extremum is hori-
zontal.
Just as the vanishing of the first derivative of a function in one variable does
not guarantee a maximum or a minimum, the vanishing of the gradient does
not guarantee a local extremum either. That is, the converse to Theorem
13.8.1 is not true in general.

Example 13.8.1
Locate and classify the critical points of f (x, y) = x2 − y 2 .

Solution.
The gradient of f is given by ∇f (x, y) = 2x~i − 2y~j. We see that fx (x, y) = 2x
and fy (x, y) = −2y are simultaneously zero at (0, 0). Therefore, (0, 0) is a
critical point and a possible extremum. The graph of f (x, y) shown in Figure
13.8.2 indicates that (0, 0) is neither a local maximum nor a local minimum.
Such a point will be called a saddle point

2
 Figure 13.8.2

We need to be able to determine whether or not a function has an extreme

value at a critical point. The following test is analogous to the Second Deriva-
tive Test for functions of one variable.

The Second Derivative Test

Let f (x, y) have continuous second derivative in an open region contain-
ing the point (a, b). This guarantee that fxy (x, y) = fyx (x, y). Suppose that
fx (a, b) = fy (a, b) = 0. Furthermore, let
 
fxx fxy 
D(a, b) =   = fxx (a, b)fyy (a, b) − [fxy (a, b)]2 .
fyx fyy 

1. If D > 0 and fxx (a, b) > 0, then f (x, y) has a relative minimum at (a, b).
2. If D > 0 and fxx (a, b) < 0, then f (x, y) has a relative maximum at (a, b).
3. If D < 0 then f (x, y)) has a saddle point at (a, b).

Proof.
First recall that the sign of f~u (x, y) determines whether f (x, y) is increasing or
decreasing in the direction of a unit vector ~u. Likewise, the sign of [f~u ]~u (x, y)
determines the concavity of f (x, y) in the direction of ~u.
Let ~u =< h, , k > be a unit vector. Then the directional derivative of f at
(x, y) in the direction of ~u is

f~u (x, y) = fx (x, y)h + fy (x, y)k.

3
The directional derivative of this new function in the direction of ~u is
∂ ∂
[f~u ]~u (x, y) = [fx (x, y)h + fy (x, y)k]h + [fx (x, y)h + fy (x, y)k]k
∂x ∂y
=fxx (x, y)h2 + fyx (x, y)hk + fxy (x, y)hk + fyy (x, y)k 2
=fxx (x, y)h2 + 2fxy (x, y)hk + fyy (x, y)k 2 .

Thus,
[f~u ]~u (a, b) = fxx (a, b)h2 + 2fxy (a, b)hk + fyy (a, b)k 2 .
Now, consider the single variable quadratic function

g(t) = fxx (a, b)t2 + 2fxy (a, b)t + fyy (a, b).

The discriminant is given by

∆ = 4fxy (a, b)2 − 4fxx (a, b)fy (a, b) = −4D.

• If D > 0 then either g(t) > 0 for all t or g(t) < 0 for all t. Since fxx (a, b) > 0,
we have g(t) > 0 for all t. But

h2
 
2 2 h
k g(h/k) =k fxx (a, b) 2 + 2fxy (a, b) + fyy (a, b)
k k
=fxx (x, y)h + 2fxy (x, y)hk + fyy (x, y)k 2 > 0.
2

This is true for any ~u. Thus, the surface is concave up around (a, b). Hence,
(a, b) is a local minimum.
• If D > 0 and fxx (a, b) < 0, we have g(t) < 0 for all t. But

h2
 
2 2 h
k g(h/k) =k fxx (a, b) 2 + 2fxy (a, b) + fyy (a, b)
k k
=fxx (x, y)h + 2fxy (x, y)hk + fyy (x, y)k 2 < 0.
2

This is true for any ~u. Thus, the surface is concave down around (a, b). Hence,
(a, b) is a local maximum.
• If D < 0 then ∆ > 0 so that g(t) changes sign. That is, there exist t1 and
t2 such that g(t1 ) > 0 and g(t2 ) < 0. Let h1 = √ t21 and k1 = √ 12 . Then
t1 +1 t1 +1
h1
t1 = k1
and

k12 g(t1 ) = fxx (x, y)h21 + 2fxy (x, y)h1 k1 + fyy (x, y)k12 > 0

4
so that f (x, y) is concave up in the direction of u~1 =< h1 , k1 > . Likewise,
we let t2 = hk22 where h2 = √ t22 and k2 = √ 12 . In this case,
t2 +1 t2 +1

k22 g(t2 ) = fxx (x, y)h21 + 2fxy (x, y)h1 k1 + fyy (x, y)k12 < 0

so that f (x, y) is concave down in the direction of u~2 =< h2 , k2 > . We

conclude that (a, b) is a saddle point

Example 13.8.2
Find the local extrema and saddle points of the function

1 y2
f (x, y) = x3 − 3x2 + + xy + 13x − y + 2.
3 4
Solution.
we first find the critical points for this function. This gives us:

fx (x, y) =x2 − 6x + y + 13 = 0
y
fy (x, y) = + x − 1 = 0.
2
From the second equation we find y = 2 − 2x. Substituting this into the first
equation we find x2 − 8x + 15 = (x − 3)(x − 5) = 0. Thus, x = 3 and x = 5
so that the critical points are (3, −4) and (5, −8).
On the other hand, we have fxx (x, y) = 2x − 6, fyy (x, y) = 12 , and fxy (x, y) =
1. Hence, D(3, −4) = −1 < 0 so (3, −4) is a saddle point. Similarly,
D(5, −8) = 2 − 1 = 1 > 0 and fxx (5, −8) = 4 > 0 so that (5, −8) is a
local minimum

Example 13.8.3
Find the local extrema and saddle points of the function

f (x, y) = x3 + y 5 − 3x − 10y + 4.

Solution.
The partial derivatives give

fx (x, y) =3x2 − 3 = 0
fy (x, y) =5y 4 − 10 = 0.

5
 √
4
Solving√each equation
√ we find
√ x = ±1 and√ y = ± 2. Thus, the critical points
4 4 4 4
are (1, 2), (1, − 2), (−1, 2), (−1, − 2).The discriminant is
D(x, y) = fxx (x, y)fyy (x, y) − [fxy (x, y)]2 = 120xy 3 .
√ √ √ √
Since D(1, 4 2) = 120
√
4
8 and f √
xx (1, 4
2) = 6 >
√ 0, (1, 4
2) is a local mini-
4 4 4
mum. √ Since D(1, − √2) = −120 8 √ < 0, (1, − 2) is a saddle point. Since
D(−1, 4 √2) = −120√4 8 < 0, (−1, 4 2) is √a saddle point. Finally, √ since
4 4 4 4
D(−1, − 2) = 120 8 > 0 and fxx (−1, − 2) = −6 < 0, (−1, − 2) is a
local maximum

The second derivative test discussed above, did not cover the case D = 0. As
illustrated in the example below, the second derivative test is inconclusive in
this case. That is one cannot classify the critical point. It can be either a
local maximum, a local minimum or a saddle point.
Example 13.8.4
Let f (x, y) = x4 + y 4 , g(x, y) = −x4 − y 4 , and h(x, y) = x4 − y 4 . Show
that D(0, 0) = 0 for each function. Classify the critical point (0, 0) for each
function.
Solution.
Note that fx (0, 0) = fy (0, 0) = 0 so that f (x, y) has a critical point at
(0, 0). Since fxx (x, y) = 12x2 , fyy (x, y) = 12y 2 and fxy (x, y) = 0, we have
D(0, 0) = fxx (0, 0)fyy (0, 0)−[fxy (0, 0)]2 = 0. But the smallest value of f (x, y)
occurs at (0, 0) so that f (x, y) has a local and global minimum at (0, 0) with
D(0, 0) = 0.
Similarly, gx (0, 0) = gy (0, 0) = 0 so that (0, 0) is a critical point of g.
Moreover, gxx (x, y) = −12x2 , gyy (x, y) = −12y 2 and gxy (x, y) = 0, we have
D(0, 0) = gxx (0, 0)gyy (0, 0) − [gxy (0, 0)]2 = 0. Since g(x, y) ≤ 0, the largest
value occurs at (0, 0). That is, g has a local and global maximum at (0, 0)
with D(0, 0) = 0.
Finally, we have hx (0, 0) = hy (0, 0) = 0 so that (0, 0) is a critical point of
h. Since hxx (x, y) = 12x2 , hyy (x, y) = −12y 2 and hxy (x, y) = 0, we have
D(0, 0) = hxx (0, 0)hyy (0, 0) − [hxy (0, 0)]0 = 0. However, h(0, 0) = 0, z =
h(x, 0) = x4 > 0 and z = h(0, y) = −y 4 < 0. Hence, (0, 0) is a saddle point
with D(0, 0) = 0
Example 13.8.5
Find the shortest distance from the point (1, 0, −2) to the plane x+2y+z = 4.

6
Solution.
Let d be the distance from (1, 0, −2) to any point (x, y, z) on the plane
x + 2y + z = 4. By the distance formula,
p p
d = (x − 1)2 + y 2 + (z + 2)2 = (x − 1)2 + y 2 + (6 − x − 2y)2 .

We have
2x + 2y − 7
fx (x, y) =
d
2x + 5y − 12
fy (x, y) = .
d
Solving 2x + 2y − 7 = 0 and x + 5y − 12 = 0 simultaneously gives x = 11 6
and
5 11 5
y = 3 so that ( 6 , 3 ) is the only critical point of f. An absolute minimum
exists (since there is a minimum distance from the point to the plane) and
it must occur at a critical point so the shortest distance occurs when x = 11
6
and y = 35 , for which d = √56

7
Absolute Extrema
In real life, one is most likely interested in finding the places at which the
largest and smallest values of a function f occur in its domain.
We recall the reader that a point (a, b) in the domain of f (x, y) is called
an absolute or global maximum if f (x, y) ≤ f (a, b) for all points in the
domain of f.
If f (a, b) ≤ f (x, y) for all points in the domain of f then f (x, y) has an
absolute or global minimun at (a, b).
Optimization typically refers to finding the global maximum or minimum
of a function. If the domain of f is the entire xy−plane then we have an un-
constrained optimization; if the domain of f is not the entire xy−plane
then we have a constrained optimization.

Example 13.8.6 (Unconstrained Optimization)

Consider the function f (x, y) = x2 (y + 1)3 + y 2 . Find the global extrema of
f, if they exist.

Solution.
The first partials give

fx (x, y) =2x(y + 1)3 = 0

fy (x, y) =3x2 (y + 1)2 + 2y = 0.

This implies that the only critical point is (0, 0). Finding second partials we
have

fxx (x, y) =2(y + 1)3 fxx (0, 0) =2

fyy (x, y) =6x2 (y + 1) + 2 fyy (0, 0) =2
fxy (x, y) =6x(y + 1)2 fxy (0, 0) =0.

Since D = fxx (0, 0)fyy (0, 0) − fxy (0, 0)2 = 4 > 0 and fxx (0, 0) = 2 > 0, the
point (0, 0) is a local minimum. Since f (−3, −2) = −5 < f (0, 0) = 0 the
point (0, 0)is not a global minimum. Thus, f has no global extrema

Like functions in one variable, a function f (x, y) can have both a global
maximum and a global minimum; a global maximum but no global minima;

8
a global minimum but no global maxima; or none. So are there conditions
that guarantee that a function has a global maximum and global minimum?
In single variable calculus we saw that a function f (x) continuous on a closed
(i.e., including the endpoints) and bounded (i.e. of finite length) interval has
both a global maximum and a global minimum. A similar result is true for
functions of two variables. However, we need to define what we mean by
“bounded” and “closed” in 2D case.
A closed set in IR2 is one which contains its boundary and with no holes
in its interior. For example, the disk x2 + y 2 ≤ 1 is a closed set whereas
x2 + y 2 < 1 is not since the boundary, which is the circumference of the circle
x2 + y 2 = 1, is not included. Similarly, 0 < x2 + y 2 ≤ 1 is not closed since it
has a hole at the origin. A bounded set in IR2 is one that can be contained
in a disk (x − a)2 + (y − b)2 < R.
Using these definitions, we have the following theorem for multivariable func-
tions:

Theorem 13.8.2 ( Extreme Value Theorem for Multivariable Functions)

If f is a continuous function on a closed and bounded set D in IR2 then f
has a global maximum and a global minimum in D.

We note that if f is not continuous or the set D is not closed or bounded,

then there is no guarantee that f will have a global maximum or minimum.
For example, the plane f (x, y) = x + y − 1 is continuous in the entire plane
but does not have global extrema since the set is not bounded.
Just as in the case of single variable functions, one can find the global extrema
by doing the following:
Step 1. Find the values of f at the critical points of f in D.
Step 2. Find the extreme values of f on the boundary of D.
Step 3. The largest of the values from steps 1 and 2 is the absolute maximum
value; the smallest of these values is the absolute minimum value.

Example 13.8.7
Find the absolute extrema of the function f (x, y) = x + y − xy in the closed
triangle with vertices at (0, 0), (0, 2) and (4, 0).

Solution.
Since D is closed and bounded and f is continuous on D, Theorem 13.8.2
guarantees that f has global extrema in D.

9
Step 1. The critical points are solutions to the equations fx (x, y) = 0 and
fy (x, y) = 0. That is, 1 − y = 0 and 1 − x = 0. The only critical point is (1, 1)
and f (1, 1) = 1.
Step 2. Along the line from (0, 0) to (0, 2), the function f (0, y) = y has
a maximum value of 2 at (0, 2) and a minimum value of 0 at (0, 0). Along
the line from (0, 0) to (4, 0) the function f (x, 0) = x has a maximum value
of 4 at (4, 0) and a minimum of 0 at (0, 0). Along the line from (4, 0) to

2
(0, 2), i.e., y = 2 − x2 , we have f (x, y) = f (x, 2 − x2 ) = 12 x − 23 + 87 for
0 ≤ x ≤ 4, a quadratic function with minimum at x = 32 , where f ( 32 , 54 ) = 87
and a maximum at x = 4, where f (4, 0) = 4.
Step 3. The absolute maximum of f on D is f (4, 0) = 4 and the absolute
minimum is f (0, 0) = 0

10